Stolz & Besner (1998) reported that there are two possible loci, morphological (i. e., letter and word) and semantic levels, for semantic priming in the interactive activation framework. To investigate the relation between levels of processing in Kanji recognition and the loci where the semantic priming occurs, event-related brain potentials (ERPs) to Kanji characters were recorded in two task conditions of semantic matching and non-task. The stimuli were presented in AAAAABAA form (habituation

more »

... paradigm) , and the repetition Kanji A and test Kanji B were semantically related, unrelated, or identical. In the non-task condition, the semantic priming effect was observed in a positive wave with a 200-ms peak latency, P200 (P2); the recovery of P200 amplitude was reduced for the semantically related test Kanji characters, compared with the unrelated ones. In the semantic matching task, where the subjects were required to decide whether or not the repetition and test Kanji characters were semantically related, the priming effect was obtained as the attenuation of the later negative wave N310 (corresponding to N400) for the related test Kanji, instead of P200. The P200 priming effect was interpreted as the semantically related response at the morphological level, which was due to semantic-level activation feeding back to the corresponding morphological-level representation, whereas the N310 effect as a reflection of the semanticlevel activation of the semantic associates with the preceding Kanji. The task dependency of semantic priming levels indicated the top-down control in Kanji recognition processes.